AI Code Review and AI Adoption – A-CX Weekly
Happy Canadian Thanksgiving to our Vancouver contingent! We’ve been writing about how one AI code reviewer replaced seven, and Mikko’s Grapevine talk on why so many companies stall with AI is now online.
AI Code Review
Ilpo has been writing about how A-CX collapsed seven AI code reviewers, whose rules had drifted apart, into one definition that runs twice, once for correctness and once for security. He covers what two independent review passes caught in the rebuild and where a person still had to make the call. Check it out here.
Measured Performance Limits
This week our development team would like to highlight work done to set performance limits from real measurements, so oversized inputs get a clear, recorded response instead of failing partway through. Limits picked on a laptop often fail under real cloud quotas and real data, and an input that is too large should be a handled case with its own message. It is worth timing your own largest realistic request against its deadline, and getting in touch if you would like a second pair of eyes on what happens when it does not fit.
Thoughts and Technology
Mikko shared some thoughts about why so many companies stall with AI in his talk at the Finnish American Chamber of Commerce North Texas. His way past the first wall is a two-week look at one real, repetitive process. Watch the talk.
Ville shared some thoughts about what coaching high school mountain bikers teaches about building with AI. His order is fundamentals, control, then speed, because teams that skip control crash faster. Read the post.
What We've Been Reading
AI has made building in-house look deceptively easy, as argued in this LinkedIn post by Charlotte Werger, who says companies underestimate the gap between a working prototype and a production system.
